5

I'm investigating dual quaternions and am having to learn a lot of stuff myself because I'm finding very few resources on the mathematical background.

I know that the magnitude of a dual quaternion is a dual number. This makes sense because it is a dual number with real quaternions jammed into its real components. Alternately, it is a quaternion with dual numbers jammed into its real components. Both approaches yield the same form.

Question: But what is magnitude of a dual number? Does that question even make sense?

I know that the magnitude of a complex number is the root of sum of the squares of the real numbers. Ditto for a real quaternion. But the dual number operator epsilon (I'll abbreviate it as simply 'e') is not like the irrational unit numbers. Square any of the irrational unit numbers and you get -1. Square e and you get 0. This makes the dual number completely useless as a rotor, but it still has use as a multi-part number. I'm wondering if nature of this dual operator also makes it unsuitable to calculate its magnitude with the Pythagorean Theorem.

Again, the magnitude of a dual quaternion is a dual number, and it doesn't make sense to me to calculate the magnitude again just to get a pure real number, so I've postulated that the magnitude of a dual number is simply itself, like saying that the magnitude of the real number 5 is simply 5. The magnitude of 1189 is 1189. The magnitude of 35 + 14e is 35 + 14e. But I don't know how to prove this or if it is even right.

Help?

The Art Of Repetition
  • 2,428
  • 1
  • 7
  • 34
John Cox
  • 171
  • 3
  • Its just the absolute value of the real part. For example $|-35 + 14\epsilon| = |-35|=35$. –  Dec 05 '14 at 00:11
  • @Bye_World why is this the right notion, out of all the possibilities one could think of? Actually, that doesn't seem to be right at all. – Kevin Carlson Dec 05 '14 at 00:12
  • It's chosen to satisfy the identity $z\bar z = |z|^2$ -- the same formula that the complex numbers and split-complex numbers satisfy. So $|-35 +14\epsilon|^2=(-35+14\epsilon)(-35-14\epsilon) = (-35)^2-(14)^2\epsilon^2 = (-35)^2$. –  Dec 05 '14 at 00:15
  • OK, I agree with this. – Kevin Carlson Dec 05 '14 at 00:20
  • 1
    That's the convention. Look up any of the works of V. Majernik on the subject. For instance this one on complex, split-complex, and dual numbers or this one on dual quaternions. –  Dec 05 '14 at 00:20
  • @Bye_World Thanks for the explanation. That makes sense. I'll have to update some of my ideas and notes that I was writing for myself, but this makes sense. – John Cox Dec 05 '14 at 00:39
  • @Bye_World Follow-up question: Where did you find the definition for magnitude of a multiple-component number? It makes sense to me, but I'd like a reference if you can provide one.

    I looked at the papers that you linked but couldn't find the words 'magnitude' or 'norm' in them.

    – John Cox Dec 05 '14 at 01:07
  • @JohnCox I don't seem to have the book anymore, but there's a nice book on this subject called Hypercomplex Numbers by I.L. Kantor & A.S. Solodovnikov. I'm relativitely sure it'll have that definition. However, look again at that first link I gave. On the second page Majernik defines this as the "absolute value" of a "two-component number system" (including dual numbers). He means the same thing you're calling the magnitude and I usually call the modulus. –  Dec 05 '14 at 14:11

1 Answers1

13

The modulus/ norm/ magnitude of a dual number $z=a+b\varepsilon$ is just the absolute value of the real part $|a|$.

Some background: Up to isomorphism, there are exactly $3$ unital $2$-dimensional algebras over $\Bbb R$. They are the complex numbers, split-complex numbers (AKA hyperbolic numbers), and dual numbers. All of them are defined to have the same formulae for conjugation and modulus. For any member of one of these algebras $z=a+b\omega$, where $\omega$ is the "imaginary" unit of the given algebra, the conjugate is defined as $\overline z = a-b\omega$ and the modulus is defined by $\|z\|^2 = z\overline z = \overline z z$.

We can see that this is a good definition, because the norm will always give us a member of $\Bbb R$: $\|z\| = \sqrt{z\overline z} = \sqrt{(a+b\omega)(a-b\omega)} = \sqrt{a^2-b^2\omega^2}$, where $\omega^2 \in \Bbb R$.

In the case of the dual numbers, $\varepsilon$ is defined by $\varepsilon \ne 0$ and $\varepsilon^2 = 0$. Then for $z=a+b\varepsilon$, we have $\|z\| = \sqrt{a^2-b^2\varepsilon^2} = \sqrt{a^2} = |a|$.

NOTE: You might also be interested to know that each of these algebras has a well-defined exponential form as well: $z=a+b\omega = \pm \|z\|e^{\operatorname{Arg}(z)\omega}$, where $\operatorname{Arg}(z)$ is akin to an angle -- in fact in the complex numbers it IS the angle. The norm of the dual number $re^{\theta\varepsilon}$ is just $|r|$.

In fact, the exponential form of the dual numbers is particularly easy. We define the exponential of a number by the infinite series $$e^x = \sum_{i=0}^\infty \frac {x^i}{i!}$$

In the case of dual numbers we have $e^{\theta\varepsilon} = 1 + \theta\varepsilon + \frac {(\theta\varepsilon)^2}2 + \frac {(\theta\varepsilon)^3}6+\cdots$. Notice that all of the terms which include a power of $\varepsilon$ greater than $1$ are $0$. Therefore this infinite series has only $2$ non-zero terms: $e^{\theta\varepsilon} = 1 + \theta\varepsilon$. Therefore the exponential form of the dual number $z=a+b\varepsilon$ is just $$z=ae^{(b/ a)\varepsilon}$$

The cavaet in this representation is clear, however: $\operatorname{Arg}(z)=\theta$ does not exist for pure imaginary numbers in the dual number plane.

  • 2
    I am not sure that we want to call this modulus a norm. Indeed, a norm requires norm z = 0 ==> z = 0. The modulus described here does not seem to respect this condition. What we have here is a seminorm, which may assign 0 to non-zero vectors. – Niriel Oct 20 '16 at 12:51
  • 1
    Note that if you choose this norm, the induced topology does not distinguish $a + b_1 \varepsilon$ and $a + b_2\varepsilon$ (and thus is non-Hausdorff); which means that the limit of the exponential series (taken under that topology) is not unique. The limit is well-defined if you use the product topology, but this isn't compatible with the norm $|a + b\varepsilon| = |a|$. – Eric Jan 17 '23 at 22:03